Vehicle Detection And Recognition Technology Research Based On TMR Sensor

With the development of China’s auto industry and the more and more vehicles caused trafficproblems become more serious in the city. In order to solve the traffic problem, people began todevelop intelligent transportation systems (Intelligent Transportation System). In the trafficinformation systems, traffic management systems, automotive systems etc of the ITS,technologyresearch vehicle presence detection, direction detection, vehicle identification, vehicle locationestimation etc were need to research. Based on TMR (Tunnel Magneto Resistance) sensor and thevehicle Information, research program was designed based on the TMR sensor vehicle detectionand recognition technology by analyzing the characteristics of the detection signal of TMR andcombining the ITS technical demands.In the vehicle detection algorithm research programs, the vehicle presence detection algorithmand vehicle direction detection algorithm based on TMR sensor was designed. Among them, whenthe vehicle presence detection algorithm was designed, vehicle signal was need to be processed.Through determining thresholds and updating thresholds, multi-state machine was used todetermine the vehicle status. When vehicle detection algorithm was designed, using the place ofTMR sensor, the results that Y, Z axes of the vehicle information can not be the reaction of vehiclestraveling direction information was known. Vehicle detection algorithm through a single sensor anddual sensor based on the X-axis direction of the vehicle was designed. Experiments showed that thevehicle presence detection and direction algorithm had a high accuracy without error detection.In the vehicle identification algorithm research programs, the GA-BP vehicle identificationalgorithm and GA-SVM vehicle identification algorithm were designed. For the shortcoming of BPneural networks, genetic algorithm was used to optimize BP neural network initial weights andthresholds. Finally, the vehicle was identified. Although the recognition algorithm achieved a92%recognition accuracy, but BP neural network exists congenital deficiencies such as easy localminimization problem. At the same time, the vehicle sample is not very enough that affecting theeffect of BP neural network recognition. SVM classification problem will be changed into quadraticoptimization problem. It`s a easy local minimization problem to solve the shortcomings of the BPneural network. The global optimal solution will be got. SVM has some advantages in dealing withthe nonlinear, small sample size problem. Therefore, GA-SVM recognition algorithm was proposedand genetic algorithm was used optimization of SVM parameters, to get the GA-SVM vehicleidentification algorithm, achieved95.3%recognition accuracy. In the vehicle location estimation algorithm research programs,vehicle location estimationalgorithm based on Kalman filter was designed. By detecting vehicle information and researchingthe magnetic signal of the vehicle trajectory curve, the magnetic signal is converted to the speed anddistance signal of vehicle. Then Kalman filter was used to estimate the relative position of thevehicle in front. Vehicle position estimation error range within0.1m.With the continuous development and expanding of ITS, the demand for vehicle detection andrecognition technology will become even more urgent.Vehicle detection and recognition technologybased on TMR sensor is a new direction.It will promote the development of ITS, make ITS play agreater role in urban construction and development.